Slashdot Mirror
FreeBSD Kernel Leak
Pine Digital Security announced a FreeBSD kernel leak, found when auditing a customer. The leak can be exploited to panic the server or elevate privileges. FreeBSD swiftly updated CVS, a security advisory will probably follow. Both the -RELEASE branch and -CURRENT branch are vulnerable.
I just wanted to point this out since whoever
submitted the article didn't feel it was
necessary.
To repeat myself, according to the article, this
problem does not effect OpenBSD or NetBSD.
*sigh* back to work...
Where in the story posting does it say that involves NetBSD and/or OpenBSD? It states clearly that it's a FreeBSD bug. And one that's already fixed in CVS to boot.
The FreeBSD security advisory is available at ftp://ftp.freebsd.org/pub/FreeBSD/CERT/advisories/ FreeBSD-SA-02%3A44.filedesc.asc, including patch.
This is a local vulnerability; it doesn't, in and of itself, make servers vulnerable. Even if someone has a local account on a system, it takes hours of CPU time to perform an exploit.
It looks like the bug (and the fix) were already announced (and committed to CVS) but that the possibility of using the bug in an exploit was not revealed until now (and might not even have been appreciated by the original reporter).
if I use FreeBSD then I will be hacked.
Not exactly a reprasentative poll but...
I use FreeBSD. I work in an office with 7 other people who all use RedHat. Out of the 8 of us, over the past 2 years, I'm the only one never to have been hacked.
The job I had before this was with an ISP which used FreeBSD for all their core systems. And in their whole history they had only ever had one FreeBSD system hacked, and that turned out to be an ex-employee who had added his public key to someobody elses authorized_keys file.
Do you mind, your karma has just run over my dogma.
I have a FreeBSD machine over at Rackspace, and I found out the hard way what they meant. They only support out-of-the-box distributions. So if you cvsup the source and then rebuild your world and kernel, then if anything goes wrong with the system (say, for instance, a disk fries) they won't perform support operations as part of your built-in service fees (since you're not running on a standard configuration). Instead, you'll have to pay a premium for the support.
Oh, go on, check out my job.
The problem isn't calling just calling fpathconf() repetitively. The problem is calling fpathconf() repetitively on a socket or other non-file (which would be a bug in itself). And by "repetitively" I mean at least 2,147,483,648 times on the same file descriptor for a system panic exploit, and exactly 4,294,967,295 times on the same file descriptor (followed by a close()) for the priviledge escalation exploit.
No network daemon that is part of the FreeBSD base system can be coerced into performing the necessary actions. Grep the source tree yourself (you'll only get a handful of hits) and examine the resulting files if you don't believe me. It's impossible to rule out everything in the ports collection (and the FreeBSD folks are careful not to make any claims regarding them) but it's hard to imagine creating an exploit of greater than theoretical importance using any network server.